The Application of Neural Networks in the Prediction of Spring-back in an L-Shaped Bend

The purpose of this work is to establish an effective prediction of the spring-back of material during the processing of an L-shaped bend. FEM-simulation of an L-shaped bend is carried out for various thicknesses of material, various punch-round-radii and die-round-radii. The spring-back depends on the shape of the bend-die and the mechanical properties of the material. The results of spring-back from FEM-simulation are then input to a neural network to establish a model for the L-shaped bend variables. The neural network is composed of a number of functional nodes. Once the L-shaped bend para-meters (material thickness, punch-round-radius and die-round-radius) are given, the bend processing performance (spring-back) can be accurately predicted by this developed network. A simulation annealing (SA) optimisation algorithm with a performance index to obtain a perfect L-shape can search for the optimal bend processing parameters. A satisfactory result was achieved based on a demonstration of simulation and on practical experience, showing that this is a new and feasible approach for use in the control of spring-back of materials.